CN111127330B - Single remote sensing image micro-ringing blocking quick wiener restoration method - Google Patents
Single remote sensing image micro-ringing blocking quick wiener restoration method Download PDFInfo
- Publication number
- CN111127330B CN111127330B CN201910982669.8A CN201910982669A CN111127330B CN 111127330 B CN111127330 B CN 111127330B CN 201910982669 A CN201910982669 A CN 201910982669A CN 111127330 B CN111127330 B CN 111127330B
- Authority
- CN
- China
- Prior art keywords
- image
- remote sensing
- hamming window
- restoration
- ringing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 230000000903 blocking effect Effects 0.000 title claims abstract description 14
- 238000012545 processing Methods 0.000 claims abstract description 24
- 238000001914 filtration Methods 0.000 claims abstract description 21
- 238000004364 calculation method Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 38
- 238000010276 construction Methods 0.000 claims description 11
- 238000009499 grossing Methods 0.000 claims description 6
- 238000013213 extrapolation Methods 0.000 claims description 2
- 238000012805 post-processing Methods 0.000 claims 2
- 230000000007 visual effect Effects 0.000 abstract description 8
- 230000000694 effects Effects 0.000 abstract description 5
- 230000015556 catabolic process Effects 0.000 description 7
- 238000006731 degradation reaction Methods 0.000 description 7
- 238000003384 imaging method Methods 0.000 description 6
- 238000001228 spectrum Methods 0.000 description 3
- 230000001627 detrimental effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000010191 image analysis Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/77—Retouching; Inpainting; Scratch removal
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/141—Discrete Fourier transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Discrete Mathematics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Image Processing (AREA)
Abstract
Aiming at the problem that the blocking restoration of a large-scale single remote sensing image generates micro ringing on the edge, the invention provides a quick restoration algorithm for eliminating the micro ringing effect of the blocking processing boundary of the large-scale remote sensing image and improving the visual effect of the image. The algorithm firstly carries out center symmetry outward expansion on the image blocks after the large-scale remote sensing image is segmented; then constructing a circular Hamming window to carry out convolution calculation on the expanded image, and covering a partial area of the convolution image with the image to be processed to obtain an edge-processed remote sensing image block; and finally, carrying out wiener filtering on the edge-processed image block, and taking the partial area of the filtered image as the restoration result of the current image block. After the method is adopted to process the large-scale single remote sensing image, the calculation amount of the algorithm is obviously reduced, the algorithm efficiency is obviously improved, the phenomenon of micro ringing of the boundary is eliminated, and the restoration visual effect of the large-scale remote sensing image is improved.
Description
Technical Field
The invention relates to the field of remote sensing image processing, in particular to a method for quickly restoring a single remote sensing image with a large picture.
Background
During the process of capturing, transmitting and storing images, the acquired images inevitably introduce different degrees of blurring and noise due to the limitation of environmental conditions and the physical limitation of imaging equipment, so that the image quality is reduced, which is called image degradation. Image degradation is quite detrimental to further applications of the image, such as feature extraction, automatic object recognition and image analysis. It is therefore necessary to improve the imaging environment and to use better imaging equipment to obtain higher quality images in order to meet the application requirements, but this is costly or even impossible under many conditions. In this case, an image restoration technique has been developed. Image restoration uses some a priori knowledge of the image to reconstruct the original image to improve the sharpness of the image and to eliminate noise. In the field of image processing, image restoration is always one of the most important and basic research subjects, and has important theoretical value and practical significance.
The large remote sensing image obtained from the satellite is affected by various factors, such as relative movement of the camera and the image, mechanical vibration of the camera, rotation and overturn of the aircraft, etc., which can result in degradation of the image quality. The degradation of the image quality is quite detrimental to further applications of the image such as feature extraction, automatic object recognition and image analysis. It is therefore necessary to improve the imaging environment and to use better imaging equipment to obtain higher quality images in order to meet the application requirements, but this is costly or even impossible under many conditions. In this context, image restoration techniques have evolved. Image deconvolution restoration techniques are one of the most common types of image restoration methods. In general, image quality degradation can be represented by using an image linear invariant convolution model, so restoration of an image using deconvolution techniques is a very effective approach.
The purpose of deconvolution is to eliminate image degradation due to the influence of convolution factors, thereby obtaining an approximate estimate of the original image. However, due to the inherent discomfort of deconvolution problems and the noise contained in the observed image, it is not feasible to directly solve the inverse of the convolution matrix formed by the point spread function. Solving the discomfort problem by the Tikhonov of the Su mathematicist gives a solution, and most of the existing image deconvolution restoration algorithms adopt the Tikhonov regularization idea and method.
R.hunt proposed constrained least squares algorithm (CLS) for image restoration in 1973. CLS introduces deterministic prior information to ensure that the norm square of the second derivative of the solution is minimal. At this time, the conversion of the image linear degradation model into Lagrange multiplier law becomes an unconstrained minimization problem. Wiener recently proposed Wiener filtered image restoration methods in 1942. Wiener filtering is a simple, general and well-effective deconvolution restoration technique. The regularized inverse filter is a regularized inverse filter and can eliminate the pathological problem caused by the frequency domain singularity of the kernel function. The greatest advantage of wiener filtering is that it is computationally efficient. The disadvantage is mainly that in order to suppress noise, the solution estimates it gives often appear too smooth. Furthermore, its assumption that both the input image and the noise are broadly stationary is often different from physical facts. At present, the problems of the restoration of a large-scale single remote sensing image are as follows: too large an image results in low algorithm efficiency and high operation cost; severe boundary ringing occurs during the blocking process, and the visual effect of the restored image is greatly affected.
Disclosure of Invention
The invention aims to solve the technical problems that: aiming at the problems of large consumption of operation and boundary ringing in block processing of a large-scale degraded remote sensing image, the micro-ringing block rapid wiener restoration method is high in operation speed, low in operation consumption and good in image visual effect.
The invention solves the technical problems by adopting the following technical scheme:
(1) Image blocking
As shown in FIG. 1, when the block FFT calculation is performed on the large-scale remote sensing image, the complexity of the integral FFT calculation can be effectively reduced, and the calculation efficiency of wiener filtering can be improved. Therefore, in order to improve the recovery efficiency of the wiener filtering image, the large-scale remote sensing image is segmented according to a proper size.
(2) Round Hamming window construction
The algorithm is mainly to extrapolate images according to a circular Hamming window construction rule, so that significant boundary ringing is avoided when wiener filtering is realized. Calculating one-dimensional Hamming window coefficients in the row and column directions according to a formula, dividing the one-dimensional Hamming window coefficients into two cases, wherein one is that a point spread function is square, constructing 7*7 Hamming windows, dividing part of neighborhood pixels into three classes according to the distance between the Hamming window and a central pixel, respectively assigning the first, second and third classes of matrix elements in the same way, and assigning the other elements of the matrix in the same way to be zero; in another case, the point spread function is not centrosymmetric, a 7*9 circular Hamming window is used to construct 9*9 Hamming window matrix, and the first, second and third matrix element assignment is the same as the first case, but the first, second and third matrix elements are slightly different, and the matrix elements in the middle of the first row and the ninth row are not listed in the fourth matrix element. The fourth class matrix elements are assigned the same value.
(3) Expanding the image to be restored, and smoothing the expanded image by using a circular Hamming window
The image to be restored is expanded along the central symmetry according to the size of the point spread function. The extended row and column images are filled with zeros. And smoothing the expanded image, namely performing convolution calculation by using a circular Hamming window as a mask to obtain a blurred image.
(4) Updating the sea window processed image and carrying out wiener filtering
And covering the blurred image obtained in the previous step by the image to be processed to form an image block after the sea window edge processing. And carrying out wiener filtering on the image block after edge processing, and extracting a single block image before edge expansion to obtain a restoration result of the image block. And then recombined according to the blocking principle to obtain a final restored image.
Compared with the prior art, the invention has obvious effects:
First, solve the great expense of the remote sensing image restoration operation of a large scale, the algorithm complexity is high, the problem of low processing efficiency. The operation cost is greatly reduced by carrying out block restoration on a large-scale remote sensing image;
And secondly, the boundary ringing effect generated by the large-scale remote sensing image blocking restoration is solved. By constructing a Hamming window, the edges of the image to be processed are processed, so that the visual effect of restoring the whole image is greatly improved;
Thirdly, performing processing by the algorithm to finish 9000 x 6144 large-scale remote sensing image restoration calculation amount which is about 1.76G MAC;
In a word, the invention is used for carrying out block restoration on a large-scale remote sensing image, firstly, a circular Hamming window is used for processing, and then, wiener filtering restoration is carried out. Therefore, the algorithm efficiency is improved, and meanwhile, the problem of boundary ringing caused by the simple image blocking restoration is solved. The image quality is obviously improved after restoration, and the algorithm efficiency is improved.
Drawings
FIG. 1 is a comparison of two-dimensional Fourier transform operational complexity under different partitioning methods;
FIG. 2 illustrates the boundary ringing phenomenon of block wiener filtering;
Fig. 3, 7 circular hamming window matrices;
Fig. 4 9 x 9 circular hamming window matrix;
FIG. 5 illustrates a micro-ringing segmented wiener filtering fast image restoration algorithm flow;
Fig. 6 shows graphs of various edge processing suppression wiener filter boundary ringing results, respectively.
Detailed Description
The invention has good effect on the restoration of large-scale remote sensing images and is suitable for engineering application. The invention carries out block processing on the large-scale remote sensing image, reduces the resource consumption in the processing process, reduces the operation complexity and improves the operation efficiency. Meanwhile, the boundary ringing effect caused by simple block restoration is avoided, and the visual effect of the restored image is influenced. The method is particularly suitable for processing large-scale remote sensing images produced in batch.
The method comprises the following specific steps:
(1) Image blocking
Dividing according to the size of the processed remote sensing image and square image blocks, and expanding the size of the divided image blocks with insufficient edges by using original boundary pixels of the image to ensure that the original boundary pixels meet the requirement of dividing the image blocks in an integer manner;
(2) Construction of circular Hamming Window
Assumed point spread functionSize/>The construction method of the circular Hamming window w (x, y) is as follows: calculating one-dimensional Hamming window coefficients in the row and column directions by adopting the method (1)
(1)
Respectively take in the row and column directionsAnd/>，/>。/>And/>The non-negative integer is taken. When/>When the method for constructing the circular Hamming window matrix is used, 7*7 is used for describing the construction method, as shown in figure 3. Dividing part of neighborhood pixels into three classes according to the distance between the first class matrix element and the center pixel element, and uniformly assigning the first class matrix element, the second class matrix element and the third class matrix element as follows:
(2)
(3)
(4)
the other elements of the matrix are uniformly assigned with zero values;
When (when) When using/>A circular hamming window matrix illustrates its construction method. As shown in fig. 4. First build/>A circular hamming window matrix. Classification of first, second and third class matrix elements/>Under the same condition, analogizing by the formula (1), and uniformly assigning the first, second and third types of matrix elements as follows:
(5)
(6)
(7)
Only the matrix elements of the fourth class are slightly different, and the matrix elements right in between the first row and the ninth row will not be listed in the matrix elements of the fourth class. The fourth matrix element is uniformly assigned as follows:
(8)
(3) Image block edge processing
Image block to be processedExtrapolated to/> along central symmetryZero padding on the edge of the size and expanding two rows and two columns to obtain/>. Image/>Smoothing the image by using the circular Hamming window constructed in the step (2), namely convoluting by using the circular Hamming window as a mask to obtain a blurred image/>. Image block to be processed/>Covering the blurred image area/>The image obtained thus far/>The image block is subjected to edge processing;
(4) Wiener filtering
The basic idea of wiener filtering algorithm (WIENER FILTER) is to find an imageIs one estimate/>So that/>And/>The mean square error between (in the statistical sense) is minimum, namely, the condition is satisfied:
(9)
Wherein, Representing mathematical expectations;
Using wiener filtering must assume that both the image and noise are a generalized stationary random process, the noise mean is zero and uncorrelated with the image, and the imaging system is a linear spatially invariant system. When using a discrete fourier transform to calculate a restoration estimate of a degraded image, the estimation formula for Wiener filtering is:
(10)
Wherein, 、/>And/>Discrete Fourier transforms of undegraded image f (x, y), point spread function h (x, y), and degraded image g (x, y), respectively,/>Is/>Complex conjugate of/>And/>The power spectra of the image and noise, respectively.
In many practical applications, the image power spectrum is due to the unaware of undegraded images and noiseAnd noise power spectrum/>It is difficult to estimate, often using a normal number γ instead of/>Gamma is typically numerically the inverse of the degraded image signal-to-noise ratio. At this time, the approximate formula of wiener filtering is:
(11)
the invention will be further described with reference to the accompanying drawings
1. Overview of the rehabilitation method
When the large-scale remote sensing image is subjected to block restoration processing, the complexity of overall restoration can be effectively reduced, and the restoration processing efficiency is improved. However, when the pixel gray values at both sides of the image block are not matched, severe boundary ringing is introduced to the image, which greatly affects the visual effect of the restored image. FIG. 1 is a schematic view ofSize image utilization/>Template wiener filtering the restored image. In order to conveniently and rapidly carry out block restoration processing on an image without causing boundary ringing, the invention provides a micro-ringing block rapid wiener restoration algorithm. According to the invention, the image extrapolation is carried out according to the proposed circular Hamming window construction rule, so that quick and effective image restoration is realized.
2. Algorithm flow block diagram
Referring to fig. 5, the remote sensing image is usually very large, the restoration operation amount for a single remote sensing image is very large, and the algorithm efficiency is low. Experiments show that when the large-scale remote sensing image is subjected to block restoration processing (shown in fig. 1), the complexity of overall restoration can be effectively reduced, and the restoration processing efficiency is improved. However, in the image blocking processing, micro ringing phenomenon occurs at the edges of the blocked images, and a large image is generated on the visual effect of the restored images. The method comprises the steps of expanding an image according to the size of the image, assigning zero to the filled edge, performing convolution calculation on the image by using a constructed Hamming window, covering a certain area of the image after the convolution calculation with the image to be processed, recovering the wiener filtering of the image block, and taking out the area with the same size as the original image in the middle as a final processing result.
3. Step (a)
The micro-ringing blocking rapid wiener restoration algorithm of the single remote sensing image comprises the following steps:
(1) Will degrade the image Extrapolated to/> along central symmetryThe size, the increased upper and lower m rows and left and right n columns are zero-filled to obtain an image/>；
(2) Image is formedMiddle/>Is assigned a value of zero;
(3) Image pair using circular Hamming window Smoothing, i.e. calculating/>Obtaining a blurred image；
(4) Using original blurred imagesCoverage area/>；
(5) By usingFor updated/>Wiener filtering is carried out to obtain a filtered image/>；
(6) Taking a filtered imageAs a final result image.
4. Application of
The method can improve the restoration speed of the single remote sensing image and improve the algorithm efficiency. Can be used for quickly recovering single remote sensing images.
Claims (4)
1. A method for quickly recovering micro ringing and blocking of a single remote sensing image is characterized by comprising the following steps: dividing according to the size of the processed remote sensing image and square image blocks, and expanding the size of the image blocks divided by the insufficient edges by using original boundary pixels of the image to ensure that the original boundary pixels meet the requirement of dividing the image blocks in an integer manner to obtain the image blocks to be processed; Circular Hamming window construction is based on the point spread function/>Extrapolation of images according to formula/>Calculating the coefficient of a one-dimensional Hamming window in the row and column directions, wherein one is that the point spread function is square, constructing 7*7 Hamming window divides partial neighborhood pixels into three classes according to the distance from the center pixel, and the first, second and third classes of matrix elements are respectively expressed as/>，，/>Uniformly assigning values, wherein other elements of the matrix are uniformly assigned to be zero; if the point spread function is not centrosymmetric, the construction of 7*9 circular Hamming window is to firstly construct 9*9 Hamming window matrix, and the first, second and third matrix elements are assigned as same as the first case, but the fourth matrix element is slightly different, according to the formulaAssigning values, wherein matrix elements in the middle of the first row and the ninth row are not listed in the fourth matrix element; expanding the image to be restored along the central symmetry according to the size of the point spread function, and filling the expanded images of each row and each column with zero; smoothing the expanded image, namely performing convolution calculation by taking a circular Hamming window as a mask to obtain a blurred image/>With the image to be processed/>Covering the blurred image area/>, obtained as described aboveForming a post-processing image block/>, of a hamming window edge; Adoption/>Post-edge processing/>Wiener filtering is carried out to obtain a filtered image/>Extract filtered image/>The restoration result of the block image is obtained; then, recombining according to a blocking principle to obtain a final restored image; and the method can improve the algorithm efficiency, simultaneously eliminate the boundary ringing problem caused by the simple image block restoration, and improve the restored image quality and algorithm efficiency.
2. The method for rapid wiener restoration of a single remote sensing image micro-ringing block according to claim 1, wherein the method comprises the following steps: the circular Hamming window is constructed according to the point spread function, for example, one is square, the construction 7*7 of the Hamming window divides partial neighborhood pixels into three classes according to the distance from the central pixel, and the first, second and third classes of matrix elements are respectively calculated according to the formula，/>， />Uniformly assigning values, wherein other elements of the matrix are uniformly assigned to be zero; if the point spread function is not centrosymmetric, the construction of 7*9 circular Hamming window is to firstly construct 9*9 Hamming window matrix, and the first, second and third matrix elements are assigned as same as the first case, but the fourth matrix element is slightly different, according to the formula/>Assigning values, the matrix elements directly intermediate the first row and the ninth row will not be listed in the fourth type of matrix element.
3. The method for rapid wiener restoration of a single remote sensing image micro-ringing block according to claim 1, wherein the method comprises the following steps: the smoothing process is to smooth the image by using the constructed circular Hamming window, that is, to carry out convolution calculation by using the circular Hamming window as a mask to obtain a blurred imageWith the image to be processed/>Covering the blurred image area obtained aboveForming a post-processing image block/>, of a hamming window edge。
4. The method for rapid wiener restoration of a single remote sensing image micro-ringing block according to claim 1, wherein the method comprises the following steps: using point spread functionsPost-edge processing/>Wiener filtering is carried out to obtain a filtered image/>Extract filtered image/>The restoration result of the block image is obtained; and the method can improve the algorithm efficiency, simultaneously eliminate the boundary ringing problem caused by the simple image block restoration, and improve the restored image quality and algorithm efficiency.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910982669.8A CN111127330B (en) | 2019-10-16 | 2019-10-16 | Single remote sensing image micro-ringing blocking quick wiener restoration method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910982669.8A CN111127330B (en) | 2019-10-16 | 2019-10-16 | Single remote sensing image micro-ringing blocking quick wiener restoration method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111127330A CN111127330A (en) | 2020-05-08 |
| CN111127330B true CN111127330B (en) | 2024-05-07 |
Family
ID=70495332
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910982669.8A Active CN111127330B (en) | 2019-10-16 | 2019-10-16 | Single remote sensing image micro-ringing blocking quick wiener restoration method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111127330B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109035139A (en) * | 2018-06-28 | 2018-12-18 | 长光卫星技术有限公司 | A kind of high-resolution satellite image modulation transfer function compensation method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7408345B2 (en) * | 2006-02-06 | 2008-08-05 | The Board Of Trustees Of The Leland Stanford Junior University | Generalized MRI reconstruction with correction for multiple image distortion |
-
2019
- 2019-10-16 CN CN201910982669.8A patent/CN111127330B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109035139A (en) * | 2018-06-28 | 2018-12-18 | 长光卫星技术有限公司 | A kind of high-resolution satellite image modulation transfer function compensation method |
Non-Patent Citations (3)
| Title |
|---|
| 基于边缘检测的邻域加窗图像去噪算法;宫霄霖;毛瑞全;刘开华;;光电工程;20091115(11);全文 * |
| 基于预估点扩展函数的光学遥感图像去模糊研究;成飞;;上海航天;20160625(03);全文 * |
| 运动模糊图像边缘振铃效应的抑制方法;王靖;施刚;张磊;韩斌;黄子娟;;电子测量技术;20130515(05);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111127330A (en) | 2020-05-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ng et al. | A total variation regularization based super-resolution reconstruction algorithm for digital video | |
| Lai et al. | Blur kernel estimation using normalized color-line prior | |
| CN102326379B (en) | Method for removing blur from image | |
| CN110675347B (en) | Image blind restoration method based on group sparse representation | |
| Su et al. | Spatially adaptive block-based super-resolution | |
| US20140355901A1 (en) | Algorithm for minimizing latent sharp image cost function and point spread function cost function with a spatial mask in a regularization term | |
| CN104103050B (en) | A kind of real video restored method based on local policy | |
| CN107133923B (en) | Fuzzy image non-blind deblurring method based on adaptive gradient sparse model | |
| CN109671029B (en) | Image denoising method based on gamma norm minimization | |
| WO2014074138A1 (en) | Globally dominant point spread function estimation | |
| Zheng et al. | Wavelet based nonlocal-means super-resolution for video sequences | |
| Sroubek et al. | Decomposition of space-variant blur in image deconvolution | |
| CN109767404B (en) | Infrared image deblurring method under salt and pepper noise | |
| CN110599402A (en) | Image super-resolution reconstruction method based on multi-feature sparse representation | |
| CN110874827A (en) | Turbulent image restoration method and device, terminal equipment and computer readable medium | |
| Xu et al. | Image deblurring with blur kernel estimation in RGB channels | |
| CN106846250B (en) | Super-resolution reconstruction method based on multi-scale filtering | |
| CN115345791A (en) | Infrared image deblurring algorithm based on attention mechanism residual error network model | |
| CN115358953A (en) | Turbulence removing method based on image registration and dynamic target fusion | |
| Umer et al. | Deep super-resolution network for single image super-resolution with realistic degradations | |
| Laroche et al. | Deep model-based super-resolution with non-uniform blur | |
| CN111127330B (en) | Single remote sensing image micro-ringing blocking quick wiener restoration method | |
| Sadaka et al. | Efficient super-resolution driven by saliency selectivity | |
| CN106033595B (en) | Image blind deblurring method based on local constraint | |
| Sundar et al. | Design and analysis of fusion algorithm for multi-frame super-resolution image reconstruction using framelet |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |